The San Andreas fault has long been viewed as a vertical structure that extends from the ground surface to the base of the lithosphere and separates the Pacific plate from the North American plate along a transform plate boundary. Analysis of Neogene structures in the central Coast Ranges indicates, however, that crustal blocks defined by strands of the San Andreas fault system have undergone pervasive compressive deformation that has shifted these blocks and their boundary faults east, relative to North America. This suggests that the blocks and their boundary structures are bounded below by an active decollement that appears to coincide with the midcrustal brittle to ductile transition at the base of the seismogenic zone. These compressive structures may be related to a currently seismically active fold and thrust belt that characterizes the eastern front of the Coast Ranges. Seismic profiles from the Central Valley, east of the range front, show that tectonic wedges (“blind thrusts”) developed within stratified rocks are present throughout this zone of deformation, and that crystalline basement that underlies these stratified rocks is not involved in the deformation. Crystalline basement may underlie most of the Coast Ranges, as indicated by the presence of metamorphic rocks of possible Sierran affinity preserved in Neogene thrust sheets as far west as Loma Prieta and by seismic velocities in the lower crust appropriate for mafic to intermediate granitoid rocks. On the basis of these relations, we suggest that the San Andreas fault system is confined to the brittle crust, above the decollement, and does not penetrate to the base of the lithosphere. Thus the San Andreas fault should not be considered as a “plate boundary.” Instead, the functional plate boundary within the Coast Ranges appears to be the inferred subhorizontal midcrustal decollement. This decollement corresponds approximately to the 350°C isotherm, the temperature at which quartz becomes ductile. If our hypothesis is proven to be correct by geologic and geophysical investigations currently underway, then extensive revisions in popular plate tectonic models applied to the Coast Ranges will be required. In addition, assessment of seismic hazards will be rendered more difficult, but more realistic, owing to realization of the greater likelihood for occurrence of thrust faulting on blind or buried faults or on faults currently deemed inactive.